How Beppe Grillo created a video in which he speaks Chinese

The effect is impressive. “Congratulations, you speak Chinese perfectly” or “I’m told it’s a Southern Chinese accent, some of the words are a little strange, but generally good.” And obviously for those who do not know a word of this language, all this seems plausible. During his last public appearance, Beppe Grillo only spoke Chinese. Founder Five Star Movementself-proclaimed Highposted a video on his blog speak fluent Chinese explaining his proposal for Via del Basilico.

However, there is a trick. The video doesn’t look like a fetus advanced chinese course or a premium subscription to Duolingue, but just good software based on artificial intelligence. The operation will be the same on the base deepfake videothose where an image of a person is used to say something.

FaceSwap Technique

Eat various techniques create a deepfake. And there are different programs. Basically it all depends on two factors: how much you want to spend and what your video editing skills are. One of the easiest ways to use is the method face change. This procedure involves replacing our face with the face of another person captured on video. When using Face Swap, our image is simply superimposed on what is in the video. Thus, any expression on your face and any word you speak will be played smoothly. absolutely true.

To achieve this goal, there are both network softwarenot very efficient, libraries on GitHub. In practice, these are already written programs that should be used by those who have at least a little knowledge of programming languages. The results are decent, in one of the tutorials found on YouTube, we can see the face Charlize Theron perfume ad video Dior J’ador Absolu replaced by the one that Mr. Bean.

How to create a deepfake video

The video of Beppe Grillo speaking Chinese seems to be taking a more difficult path. at least two steps. The first was create an audio track with the voice of Beppe Grillo who accentuates his speech in Chinese. Cloning an entry is relatively easy. Example. VALL-E, an artificial intelligence software developed by Microsoft, allows you to clone any voice in just three seconds. To say what we want, it is enough to insert the text already translated into Chinese.

As soon as we have audio track you will have to apply this to the video. There are also many roads here. Software unearth promises to transform our content into any language. However, the result is a bit cheesy. Already in commercials it is clear that the sound is not synchronized with the movements of the mouth. Other programs such as Model of spline motion of a thin plate allow you to apply sound to a static image that starts moving and talking. Here is the result pretty impressive: during speech, not only the lips move, but the whole face changes and moves.

When Video by Beppe Grillo it looks like a different technology was used. Here, in fact, the sound is directly connected to the video. Here, too, there is a lot of software, both the one that allows you to do everything online, and the one that requires a little learning. After some tweaks and proper video editing, everything works. On the other hand, now I deepfake video they also became a regular feature even Strip the news.